Wancong (Kevin) Zhang

I am a second year PhD student at NYU Courant, specializing in deep learning under the guidance of Yann LeCun.

Prior to my PhD, I worked as a senior researcher at AssemblyAI, where I led the developments of Conformer-1, Conformer-2, and realtime automatic speech recognition (ASR) models. I earned my M.S. degree from NYU Courant Institute's Computer Science Department, where I had the privilege to collaborate with Nicolas Carion, Marzyeh Ghassemi, and Rajesh Ranganath. Before starting my AI journey, I explored the world of stem cell and molecular biology research at Harvard Stem Cell Institute.

Email  /  CV  /  Google Scholar  /  Github

Research (Updated 07/2025)

I believe the ability to learn from observations and interacting with the world in an unsupervised manner is a crucial stepping stone to unlocking general intelligence. My primary research focus revolves around self supervised learning (SSL) and novelty guided explorations, with an emphasis on their applications in computer vision and control.

This work introduces zero-shot planning using the Joint Embedding Prediction Architecture (JEPA) trained from offline trajectories, demonstrating its strengths in generalization, trajectory stitching, and data efficiency compared to traditional offline RL.

Presents an efficient probing benchmark to evaluate the fitness of unsupervised visual representations for reinforcement learning (RL). Applied it to systematically improve pre-existing SSL recipes for RL.

Showcases an industrial-scale end-to-end Automatic Speech Recognition model trained on 570k hours of speech audio data using Noisy Student. It achieves competitive word error rates against larger and more computationally expensive models.

Adapts the computer vision data augmentation technique MixUp to the natural language domain, reducing calibration error of transformers for sentence classification by up to 50%.

Establishes a pre-training benchmark protocol for electronic health record (EHR) data.

Teaching